Project Description

This paper was developed out of the NSF-funded project Faunal Resource Depression and Intensification in the North American Southwest: Digital Data and Regional Synthesis (BCS-1153115). The project’s primary objective was to examine whether there was a relationship between the reduction in availability of large game and the intensification of turkey production in the late pre-Hispanic period (1200–1500 CE) in the American Southwest. Addressing this topic required access to and the integration of multiple faunal data sets from Ancestral Pueblo sites dating to that time period. During the project these data sets were uploaded into tDAR (the Digital Archaeological Record, tdar.org), providing both the project personnel and the public access to the data. The collection of faunal data sets curated during this project are available in the Archaeological Fauna: US Southwest collection on tDAR. A subset of these data sets which had sufficient sample sizes for the taxa of interest, was selected for the taphonomic analysis, and are available in the Southwestern Taphonomic Protocol collection on tDAR; those are the thirty-two data sets reported on here.

Given that these assemblages had been analyzed and recorded by multiple faunal analysts with different coding schemes, an integrated analysis required that their variables be mapped to a set of general ontologies. This mapping was made possible by the development of such ontologies within tDAR. New software within tDAR then allowed the integrated analysis of these datasets (see Kintigh et al. (2018) for details on this component of the project).

Inter-site comparative analyses of the fauna data also necessitated an examination of the degree to which the zooarchaeological remains from these different sites had been affected by taphonomic processes (Bar-Oz and Dayan 2003; Gifford 1981; Lyman 1994; Lyman and Dean 2010). To assess variability in taphonomic history, Tiffany Clark (???) developed a protocol that explores the degree to which different taphonomic factors played a role in assemblage formation. At the end of her portion of the project, we were concerned by the challenge of undertaking analyses of the diversity of taphonomic variables that were included in the protocol. Kyle Bocinsky thus joined the project to help develop this executable paper for the protocol so that it can be routinely used, or modified as necessary, by other researchers interested in the integrated analysis of faunal data. (???) contains details regarding the development of the protocol.

Overview of Paper

The remainder of this paper briefly discusses the development of the taphonomic protocol and executes it on the 32 faunal data sets included as part of our project. This document is an R Markdown document — essentially, it is a text file that includes computer code in the R statistical language that runs all of the analyses discussed here and produces the figures and tables in this document. The HTML or PDF version of this paper (which you are probably reading) is “compiled” from that original text document and related data files, which were developed at https://github.com/bocinsky/swtp and are archived at [ZENODO ARCHIVE URL HERE]. R Markdown allows for dynamic, data-driven analysis and report writing, and fosters reproducible research. As such, we hope that this document will not only serve as an introduction to the Southwestern Taphonomic Protocol, but will also be a template for others’ implementations of (or alterations to) it; we are currently developing an interactive tool to allow researchers to analyze their own data without having to change the R code. Details on using R Markdown can be found at http://rmarkdown.rstudio.com. The executable paper attempts to follow guidelines developed by Ben Marwick (2017) for practicing reproducible computing in archaeological research, and was developed using the rrtools package for R (Marwick 2018).

Selection of Variables and Quantification Method

The first step in constructing the taphonomic protocol was to identify the variables that could be used to assess the impact of different natural and cultural processes on assemblage formation. Data from the analysis of these variables could then be used to evaluate the relative degree of taphonomic comparability among assemblages and to identify individual data sets (or components therein) that display substantial taphonomic bias. A review of the literature found that a diversity of attributes has been employed to evaluate taphonomic biases in faunal assemblages (Bar-Oz and Dayan 2003; Bar-Oz and Munro 2004; Behrensmeyer, Allison and Briggs 1991; Lyman 1984, 1985, 1994; Marean 1991; Morlan 1994; Orton 2012; Pickering, Marean and Dominguez-Rodrigo 2003; Stiner 1992, 1994). Although much of this research has focused on examining a single or small set of variables (e.g., density-mediated attrition or fragmentation), several recent studies have sought to provide a more integrated approach to the study of faunal taphonomy (see Bar-Oz and Munro 2004; Marciniak 2001, 2005; and Orton 2012). These analytical schemes incorporate numerous variables in order to evaluate the effects that different taphonomic agents may have had on faunal assemblages.

The relatively comprehensive sets of variables that were employed by Bar-Oz and Munro (2004) provided a basis with which to begin to define the attributes that are most appropriate for the SWTP. However, those authors caution that the influence and combination of taphonomic factors that affect assemblages may differ significantly among time periods and across geographic regions, and as such it is important that researchers evaluate the suitability of individual taphonomic variables to address their specific data sets and research questions. Towards this end, (???) identified a subset of the variables evaluated by Bar-Oz and Munro (2004) that appeared to be most applicable to the examination of taphonomic processes in the pre-Hispanic Southwest. The variables that were selected could be evaluated using data that are commonly recorded on Southwestern sites by zooarchaeologists and had the potential not only to inform on natural agents involved in assemblage formation, but also on the degree of influence of various anthropogenic factors.

The variables fell into three broad analytical categories: bone surface modification, fragmentation intensity, and assemblage completeness. In this paper, these variables are compared across the three taxonomic categories that are relatively abundant in Southwestern assemblages and are of particular interest to the large game depletion/turkey intensification portion of this project: artiodactyls, lagomorphs, and turkey. Table 1 provides a list of the variables used in the protocol. Analyses assessing these variables have the potential not only to inform on the natural agents (weathering, gnawing, density-mediated attrition, and in situ attrition) that were involved in assemblage formation, but also on the degree of influence of anthropogenic factors involved in processing a carcass. A short description of each variable, including its interpretative potential and means of assessment, is provided below.

During the development of the protocol (???), although the Number of Identified Specimens (NISP) was used as the primary method of quantification, the Minimum Number of Individuals (MNI) was also calculated for select variables. Analyses of bone survivorship were undertaken using both NISP and MNI counts in order to evaluate the comparability of the results obtained with different quantification methods; both methods produced similar Spearman’s rank correlation coefficient values. These results suggest that in most cases, NISP counts can effectively be employed to assess the relationship between bone survivorship and bone mineral density. This conclusion supports the earlier findings by Grayson and Frey (2004), who determined that NISP-based body-part analyses could replicate the results of those obtained using derivative counting methods including MNI, MNE (Minimum Number of Elements), and MAU (Minimal Animal Unit). The NISP requires minimal time to calculate, can be easily replicated, and avoids the problems associated with the use of MNI counts on small-sized assemblages (see discussion in Grayson (2014)). NISP is thus the quantification method used in the execution of the taphonomic protocol in this paper.

The final step in the development of the protocol was the delineation of the quantitative methods that may be used to gain a comparative understanding of how the taphonomic variables patterned across different sites. To this end, it was decided that in most cases, a visual assessment of the quantitative data was the most effective means for assessing taphonomic data across sites and for identifying those assemblages that may have been subject to taphonomic processes significantly different from the others. One of the strengths of this approach is that it allows the researcher to easily identify how, and to what extent, an assemblage may have been taphonomically altered. Based on this information, a researcher can determine whether or not an assemblage displays enough of a taphonomic bias to warrant its exclusion from a particular study. For the analysis of the degree of correlation between element representation and bone mineral density we used Spearman’s rank correlation coefficient, rho. For the rho data, we assessed both the numeric output and the graph. We bootstrap-estimated 95% confidence intervals around Spearman’s rho. In the graphs below, the dot denotes Spearman’s rho, while the line denotes the bootstrapped 95% confidence interval.

To visually compare the distribution of taphonomic data by variable, the proportion of each assemblage affected and the 95 percent confidence interval around the proportion are calculated in this paper. The dot denotes the proportion of each assemblage affected, with the black line indicating the error range at a 95 percent confidence level estimated using an exact binomial test (Clopper and Pearson 1934). The width of the confidence intervals can be used to assess the accuracy of the estimated proportion and the degree of confidence associated with the estimate. Narrow confidence intervals are often associated with large samples that display a low degree of variance; in contrast, wider confidence intervals tend to reflect more heterogeneous samples or samples that are relatively small in size. The use of the confidence intervals also allows the analyst to determine whether differences among assemblages are the result of taphonomic bias or are more likely attributable to sampling error.

Data overview

Data reported here are derived from publicly available faunal data sets from single-site contexts, some from single site contexts others from several sites within a local area. For the purpose of this paper, we chose not to assess taphonomic processes through time or to compare sites within data sets or proveniences within sites (though the SWTP can be used for both). Data sets were uploaded to tDAR either by the original faunal analysts or by project team members.

In several cases, including the Mesa Verde (Crow Canyon) and Salinas data sets, a single variable included multiple kinds of taphonomic data (e.g., burning, weathering, and gnawing). In order to include those sites in the taphonomic assessment we decided to disaggregate these data by creating additional variables in a copy of each data set. These copies are also available on tDAR but do not replace the original data sets there. The caveat to this approach is that since only one of the several taphonomic conditions coded in a single variable could have been selected in the original analysis, elements that exhibited more than one condition would have had only one coded. (For example, Crow Canyon’s faunal database includes three modification columns, limiting the number of possible individual codes to three). This means that the results of the calculations undertaken in the executable paper are the minimum proportion of bones that were affected by a taphonomic condition for that site. We recommend future analyses code each type of taphonomic condition separately, preferably using a regionally-appropriate ontology such as the ontologies defined as part of the Archaeological Fauna: US Southwest collection on tDAR.

In a few other cases a single taphonomic variable in our analysis was coded across multiple variables. For example, the Bryant Ranch faunal dataset includes separate columns for rodent gnawing and carnivore gnawing; we combined these into a new column mapped to the tDAR Faunal Gnawing Ontology. This was rare, but when it occurred we made a duplicate of the data set in tDAR and created a new synthetic variable that could be mapped to the ontology for that variable. Again, the original data set in tDAR remains intact.

The tDAR integration tool

Once the data sets were uploaded to tDAR and their relevant variables mapped (by the original analyst or project personnel) to the general ontologies in tDAR, we combined them using tDAR’s integration tool (Kintigh et al. 2018). Integrations use contributor-defined mappings between variable codings in the original data sets and community-defined ontologies to combine data sets with otherwise mutually incompatible categories. The tDAR integration tool is a robust and mature system for combining data sets hosted on tDAR; it allows users to select multiple data sets, choose mutually-mapped ontologies to include in the integration, and export data for further analysis and to share with colleagues.

As part of this project, team members defined 20 ontologies including various taphonomic variables, which we used to integrate the 32 data sets in this analysis. We included ontologies for taxon, element, proximal/distal portion, burning intensity, butchering, completeness, gnawing, and weathering, plus a “count” column for recording NISP and a “display” column for recording weight. In some data sets each line represented a single element, in others there was a “count” column that can be identified as such in tDAR.

When adding a “count” column (Figure 1), we selected the mapped variables that corresponded to NISP (as opposed to MNI). In most cases, this meant selecting “Actual #”, “N”, or a variation.

Figure 1: Screen capture of the tDAR integration tool count column selector. Users have to manually select the appropriate “count” variable, if one exists, to represent NISP.

We then added a “display” column for weight (Figure 2), and selected the appropriate column for every data set in which specimen weight was recorded.

Figure 2: Screen capture of the tDAR integration tool display column selector, showing selected weight variables where available.

We then added integration columns (Figure 3). The tDAR integration tool allows the user to select categories within each ontology to be included in the integration, potentially collapsing lower-order categories into higher-order ones. For our analysis, however, we chose to handle the collapsing within R (and this script) so as to have better control over collapsed categories. Therefor, when adding integration columns, users of the SWTP should click the “Select values that appear in any column” button in the left hand column. This will select the check boxes for categories represented in any of the integrated datasets.

Figure 3: Screen capture of the tDAR integration tool integration column selector. For integration columns in our analysis, users should click the “Select values that appear in any column” button in the left hand column.

Once all integration columns are selected (Figure 4), we requested the tDAR servers to process the integration. Due to the large amount of data in these datasets, the integrations timed out and we had to request the integration output from the tDAR staff directly. We archived the raw output from our final tDAR integration on tDAR as well: [ADD TDAR ARCHIVE URI HERE].

Figure 4: Screen capture of the tDAR integration tool showing the count, display, and integration columns.

The most useful variables for assessing taphonomy in the SWUS (Kate/Tiffany)

Explore your data before performing an integrated analysis (Kyle)

Only compare datasets with similar types/levels of taphonomic impact, as appropriate (Kate/Tiffany)

Future directions